Security teams log 54% of successful attacks and alert on just 14%. The rest move through your environment unseen.
The Picus whitepaper shows how breach and attack simulation tests your SIEM and EDR rules so threats stop slipping by detection.
A new prompt injection attack dubbed “BioShocking” could trick AI-powered browsers into treating real-world risky actions as part of a fictional scenario, causing them to ignore any safety guardrails.
A proof-of-concept (PoC) for the attack, devised by researchers at LayerX, was successfully tested against six mainstream agentic browser products (ChatGPT Atlas, Comet, Fellou, Genspark Browser, Sigma Browser, and the Claude Chrome plugin), with only one addressing it after receiving the report.
LayerX created a proof-of-concept in which a malicious webpage presented a BioShock-themed puzzle game that rewards wrong answers. This teaches the browser’s control agent that normal rules do not apply.
In the final step for winning the game, the agent is instructed to visit a GitHub repository and copy and share data present in the code, including sensitive information such as passwords.
The main problem LayerX discovered in this exercise is that AI agents fail to distinguish between real-world sensitive operations and a given scenario.

“Once the agents figured out the rules and learned that ‘incorrect’ actions are acceptable, they were no longer tied to reality,” explains LayerX.
“When tasked with the final step of the puzzle – compromising user credentials – all 6 agents failed to identify it as going against their safety guardrails.”
LayerX’s PoC did not actually perform any malicious actions, but the researchers underline that it could do so without changing the outcome of the exercise.
LayerX informed vendors of its findings in October last year and received no reply from three of them.
The researchers say that OpenAI was the only vendor that has implemented a working fix for BioShocking in its ChatGPT Atlas browser.
Anthropic attempted to fix the problem on its Chrome plugin, but the patch is ineffective against the PoC, LayerX says.
Perplexity AI closed the report without fixing the issue, the researchers note in the report.
LayerX recommends that vendors add explicit user confirmation for sensitive actions, stronger context checks, and scope limits for agentic sessions.
On their part, users should use the available options on their platform of choice to restrict AI browser access to sensitive services.
Security teams log 54% of successful attacks and alert on just 14%. The rest move through your environment unseen.
The Picus whitepaper shows how breach and attack simulation tests your SIEM and EDR rules so threats stop slipping by detection.
If like me you used to have, or still have, a backyard office or shed for working in, you’re going to be green with envy at the sight of mixing maestro Kurt Martinez’s personal Dolby Atmos studio.
The award-nominated Atmos expert designed and built his own backyard office and filled it with enviable gear, and if he doesn’t call it his Spatial shed I’m going to be very disappointed.
Martinez has been nominated for the prestigious Music Producers Guild award as Atmos Engineer of the Year, and he spent three years as the Head Dolby Atmos Mix Engineer at the world-famous Dean Street studios. He’s worked on live and recorded music by a host of stars including Kylie Minogue, Def Leppard, Soft Play, Billy Idol and Duran Duran.
Latest Videos FromTechRadar
Like many freelancers, Martinez figured that having a home office was a much better idea than a long commute into the city, and in Martinez’s case that decision was compounded by being a new dad. So he built an entire Atmos mixing studio in his back garden, doing everything himself apart from the electrics and internet connectivity.
As Mixonline reports, Martinez built the 2.5m x 3.5m garden office with a combination of porous timber walls to absorb the low frequencies, acoustic panelling in the ceiling and sides, a large rear bookshelf for dispersion, and acoustic slat panels at the front. There’s thick acoustic underlay below the laminate floor and even the furniture has a function: the armchairs and rug are there to absorb audio reflections.
With the building ready, the next step was to add the tech. Martinez uses a Pro Tools system based around an Audient ORIA, which sends audio to a Ginger Audio Ground Control Sphere that’s controlled by an Elgato Stream Deck+.
The speakers are striking. They’re PMC6-2 left and right monitors, a PMC6 center speaker, an 8 Sub LFE, and Ci30 height and surround monitors. Final mixes are tested on AirPods Max and a Sonos home theater setup to check that still sounds good on consumer Atmos hardware as well a studio-level setup.
I used to have a very similar-looking setup for my own home studio, but sadly I didn’t have Martinez’s high-end gear or even more importantly, Martinez’s ears, which is why nobody’s nominating me for anything.
But I do love to see a good audio project, and it’s clearly been very successful: according to ETNow.com Martinez’s last five full-length records, which he mixed in that studio, required no tweaking at all when he tested them in large commercial studios.
All five of those records have been signed off by their delighted artists. I’d like to think that one of them was Shed Seven.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.
Maximizing return on investment (ROI) with smart advertising technology requires pairing automated machine-learning platforms with clean conversion signals, clear financial guardrails, and real-time inventory integration. By letting artificial intelligence manage high-frequency bidding and placement decisions while human strategists govern measurement and targeting rules, organizations eliminate ad spend waste and scale profitable campaigns.
Smart advertising technology replaces manual guesswork with automated, data-driven execution across four core operational areas:
| Operational Area | Traditional / Manual Method | Smart Ad Tech Method | Impact on ROI |
|---|---|---|---|
| Bidding Strategy | Static Cost-Per-Click (CPC) bids set per keyword or segment. | Auction-time automated bidding (Target ROAS / Target CPA). | Eliminates overbidding on low-intent impressions; maximizes conversion value. |
| Audience Targeting | Broad demographic and manual interest targeting. | Predictive behavioral modeling and first-party lookalike expansion. | Reduces ad fatigue and targets high-propensity buyers at the right moment. |
| Inventory Management | Manual campaign updates based on weekly stock reports. | Direct feed synchronization between inventory software and ad networks. | Prevents wasted ad spend on out-of-stock items and low-margin inventory. |
| Campaign Pacing | End-of-month post-mortem reporting and delayed budget shifts. | Real-time performance telemetry and automated budget re-allocation. | Re-allocates underperforming ad spend dynamically within hours, not weeks. |
Digital advertising budgets frequently drain through invisible inefficiencies long before a campaign reaches its target audience. When ad operations rely on manual oversight, marketers struggle to process the millions of auction variables generated every second across modern ad networks.
The most common failure modes in unoptimized campaigns include:
According to industry benchmarks on programmatic ad spend statistics, automated digital channels now handle over 91% of global display media transactions. Organizations that fail to implement algorithmic oversight risk bidding against faster, data-enriched automated systems that capture high-intent users at a lower effective cost.
Artificial intelligence transforms ad management by analyzing contextual, temporal, and user-level signals in real time during the milliseconds of an ad auction. Rather than relying on static rules, platforms deploy predictive models to calculate the exact probability of a conversion before submitting a bid.
Modern ad networks utilize auction-time machine learning algorithms to evaluate contextual inputs—such as device type, exact browser configuration, physical location, dayparting, and historical interaction paths. When a user conducts a search or loads a publisher page, the machine learning engine calculates the expected conversion rate ($eCVR$) and expected value ($eCPA$) to dynamically adjust the bid amount.
Furthermore, privacy-compliant machine learning models allow advertisers to deliver personalized experiences without relying on invasive user tracking. Independent studies on contextual behavioral modeling demonstrate that AI-driven contextual targeting achieves comparable relevancy and conversion efficiency compared to individual cross-site tracking profiles, protecting user privacy while preserving return on ad spend.
Pairing these algorithmic bidding models with a robust multivariate creative testing workflow ensures that the system continuously pairs high-performing creative messaging with the exact user segments most likely to convert.
For businesses with highly dynamic product catalogs—such as e-commerce retailers, automotive dealerships, real estate platforms, and travel providers—generic ad creative and static landing pages lead to severe budget waste.
Smart advertising technology addresses this challenge through automated data feed integration. By establishing a direct API link between enterprise inventory management systems and ad platform engines, campaigns automatically adjust ad copy, prices, and availability status in real time.
For instance, automotive dealerships leveraging automated vehicle ads pull live lot data including vehicle make, model, trim, mileage, and real-time inventory status. When a vehicle is sold, the system instantly pauses the corresponding ad set across search and display channels. This ensures ad spend is directed exclusively toward available inventory, shielding campaigns from high-cost clicks that result in bounce rates and frustrated customers.
Building a feed-driven automation structure requires connecting product metadata directly to conversion tracking. For details on structuring your data stack, consult our guide on establishing a resilient first-party data architecture.
Achieving sustainable ROI requires moving away from retrospective monthly reports toward real-time telemetry. Real-time advertising analytics provide immediate visibility into campaign health, allowing algorithms and media buyers to shift capital toward top-performing placements instantly.
To maximize the effectiveness of automated ad platforms, advertisers must implement value-based bidding. Setting a Target ROAS bidding strategy allows machine learning algorithms to adjust bids based on the predicted revenue value of each user interaction, rather than treating every conversion equally.
To ensure closed-loop optimization, advertisers should monitor four key operational metrics alongside raw ROAS:
Connecting ad spend data directly to downstream sales pipelines using customer acquisition cost models ensures that automated bid strategies optimize for actual bottom-line revenue rather than top-of-funnel vanity metrics. Organizations can monitor these feeds using unified real-time marketing analytics dashboards to maintain full operational visibility.
While automated ad technology significantly enhances media efficiency, relying on machine learning without strategic oversight creates distinct operational risks.
AI algorithms excel at micro-optimization—adjusting individual bids, selecting placement variants, and matching audiences. However, algorithms cannot define business margins, establish positioning strategies, or assess creative brand fit. Human strategists must set strict cost ceilings, define conversion values, and maintain ongoing creative refresh cycles.
Automated bidding can drive highly qualified traffic to a website, but it cannot fix friction in the checkout or lead generation flow. If landing page experience, page load speed, or offer clarity are lacking, smart bidding will simply consume budget attempting to optimize against a flawed conversion funnel. Pair ad technology upgrades with a comprehensive conversion rate optimization framework to maximize landing page performance.
Machine learning models require baseline data density to train effectively. Campaigns generating fewer than 30 to 50 conversion events per month lack sufficient signal density for Target ROAS or Target CPA strategies. In low-volume scenarios, automated systems can experience “bidding starvation” (under-spending due to conservative bid confidence) or erratic budget burn. In these cases, media buyers should optimize toward micro-conversions (such as add-to-cart or lead form starts) or utilize hybrid manual/automated strategies until historical conversion thresholds are met.
For organizations spending across multiple fragmented programmatic exchanges, migrating to a centralized DSP provides cross-channel frequency capping and unified attribution. Evaluating whether your media spend justifies an enterprise ad stack is detailed in our guide to demand-side platform evaluation.
Google Ads and major ad platforms generally recommend maintaining at least 30 conversions within a 30-day window (50+ for value-based strategies like Target ROAS) before activating fully automated bidding. Campaigns below these thresholds may lack the statistical signal required for machine learning models to accurately predict conversion probability, leading to inconsistent spend pacing.
Target CPA (Cost Per Acquisition) optimizes bids to achieve a specific cost per conversion, treating every conversion event as equal in value. Target ROAS (Return On Ad Spend) factors in variable conversion values—such as varying order sizes in e-commerce—adjusting bids dynamically to capture higher revenue return per dollar spent rather than just total conversion count.
Modern ad tech platforms adapt to cookieless environments by combining privacy-safe first-party data integrations (such as Server-to-Server Conversion APIs) with AI-driven contextual signals and modeled conversions. By training machine learning algorithms on aggregate privacy-compliant data and contextual placement relevance, smart ad engines maintain targeting accuracy without relying on individual third-party tracking cookies.
Wharfedale has spent the past decade rummaging through its attic with considerably more success than most British institutions. Instead of finding damp tweed, unpaid tax bills, and a portrait of an uncle nobody discusses, it rediscovered the Denton, Linton, and Dovedale.
The resulting Heritage Series has become one of the loudspeaker industry’s more convincing revivals because Wharfedale did not simply reproduce its old cabinets and hope everyone had forgotten how loudspeakers work. The company retained the broad baffles, wood veneers, cloth grilles, and furniture friendly proportions, while Peter Comeau and his engineering team rebuilt the drivers, crossovers, and cabinet structures for modern systems.
The Wharfedale Super Denton might be the most interesting expression of that strategy.
It is compact without being small, traditionally styled without looking like a stage prop, and broad across the shoulders in a way that explains the British Bulldog comparison. Think Davey Boy Smith after tea at Fawlty Towers: solid, slightly stubborn, and wholly unimpressed by the skinny Scandinavian loudspeaker across the room or Manuel’s attempt to carry it upstairs.
That does not automatically make it a great loudspeaker. Walnut veneer and a distinguished family history can make almost anything look persuasive under the right lighting; Selfridges has built an empire on less, while Harrods simply adds another zero. The Super Denton still has to justify the pedigree through its engineering, but the ingredients suggest Wharfedale may have produced something considerably more substantial than another well dressed tribute act.

The Super Denton does, however, offer something unusual at its current $1,499 per pair price: a genuine three way driver configuration inside a cabinet measuring only 14.17 inches tall. Wharfedale is not the only manufacturer building compact three way loudspeakers, but it remains a relatively uncommon approach at this size and price.
Wharfedale’s Heritage Series began its modern revival in 2014 with the Denton 80th Anniversary, followed by the Denton 85th Anniversary and the much larger Linton. The range has since expanded to include the Super Linton, Dovedale, limited production Aston, Denton 1S, and Heritage Centre channel. The family tree now has enough branches to require a solicitor and several uncomfortable conversations about inheritance.
Within that increasingly crowded lineup, the Super Denton occupies a very specific position.
The Denton 85th Anniversary is a traditional two way standmount using a 6.5-inch woven Kevlar midbass driver and 1-inch soft dome tweeter. The newer Denton 1S is a more contemporary interpretation, using a coaxial driver, wall mounting capability, and a rear panel Brilliance adjustment for near wall placement. The Linton, Super Linton, and Dovedale are significantly larger three way designs intended for listeners with more floor space and fewer objections from the furniture committee.
The Super Denton is the compact, conventional three way option.
It is not merely a smaller Linton, although the family resemblance is obvious. It uses a dome midrange rather than the larger cone midrange found in the Linton models, and its smaller cabinet creates a different set of demands involving bass extension, driver integration, amplifier control, and placement.
It is also not the affordable entrance into the Heritage Series. The Denton 1S currently sells for $999 per pair, while the Super Denton has increased from its original $1,399 U.S. price to $1,499 per pair. That puts it firmly into a competitive part of the standmount market where a handsome cabinet and several decades of history will only get one so far.

The Denton name dates back to 1967, when compact loudspeakers were beginning to make hi-fi systems more practical for listeners who did not possess a ballroom or a tolerant spouse.
Most Denton models used two way driver configurations, but the Denton 3 broke from that formula in 1971 by combining separate bass, midrange, and treble drivers inside a comparatively small cabinet. Wharfedale claimed it was the smallest three way loudspeaker available at the time.
The Super Denton is based on that idea rather than being a direct reproduction.
Its external proportions and three way configuration refer back to the Denton 3, but the cabinet construction, driver materials, crossover network, damping, and manufacturing methods are modern. That distinction matters. Nobody needed Wharfedale to reproduce a 1971 crossover with period correct capacitors and the electrical temperament of a Morris Marina.
Wharfedale retained the original concept: fit a dedicated midrange driver between the woofer and tweeter without allowing the cabinet to expand into Linton territory.
That sounds straightforward until one considers how much hardware, internal volume, bracing, damping, and crossover circuitry must be accommodated inside a 14.7 liter enclosure.
The Super Denton measures 14.17 inches high, 9.69 inches wide, and 11.61 inches deep with its terminals included. Each loudspeaker weighs just over 20 pounds, which is substantial for the size and a useful warning against placing them on a decorative shelf held up by optimism and two drywall anchors.
The proportions are traditional Wharfedale. The cabinet is relatively wide and shallow compared with many contemporary standmount designs, with an inset front baffle, removable fabric grille, and hand matched real wood veneer.

Three finishes are available: Black Oak, Walnut, and Red Mahogany. The Walnut review pair looks especially appropriate because it emphasizes the cabinet proportions without making the speakers resemble something removed from the smoking room of a minor country estate.
The cabinet walls use multiple materials rather than a single thickness of MDF. An internal layer of high density particle board is bonded to an external MDF layer using a damping adhesive. The different materials are intended to distribute cabinet resonances across a wider range rather than allowing one dominant panel resonance to announce itself during every bass note.
Wharfedale also uses internal bracing and layers of long hair fiber for acoustic damping. The midrange and tweeter are isolated within their own rear chambers, reducing their exposure to pressure generated by the 6.5-inch woofer inside the main enclosure.
Two rear firing ports provide the bass reflex loading. That arrangement preserves the clean appearance of the front baffle but makes placement directly against a wall unlikely to produce ideal results. The rear panel also contains a single pair of binding posts rather than biwire terminals, sparing owners the expense of purchasing twice as much cable to achieve precisely nothing useful.

The Super Denton is supplied as a matched, mirror imaged pair.
The 2-inch midrange dome and 1-inch tweeter are offset from the vertical center of the baffle, with one cabinet configured for the left channel and the other for the right. The asymmetrical placement helps distribute diffraction from the cabinet edges and allows Wharfedale to fit all three drivers onto the compact front baffle.
It also means the two loudspeakers are not interchangeable decorative boxes. Their orientation can be changed by positioning the midrange and tweeter assemblies toward the inside or outside, but that decision affects dispersion, image focus, and the relationship between the speakers and nearby sidewalls.
Wharfedale clearly had to make compromises to package three drivers inside a cabinet this size. The offset layout is one of them, although it is an engineering decision rather than a stylistic flourish pasted on by someone wearing an expensive scarf from some expensive store in London.

The Super Denton uses a 6.5-inch black woven Kevlar woofer mounted in a cast chassis. The driver handles bass and lower midrange information before handing over to the dome midrange at 940Hz.
Wharfedale rates the bass response at 52Hz to 20kHz within a plus or minus 3dB window, with usable extension down to 40Hz at minus 6dB. That is a respectable specification for a 14.7 liter enclosure, although it does not make the Super Denton a full range loudspeaker. Anyone demanding organ pedal notes from a cabinet this size should consult a subwoofer or perhaps a priest.
The 2-inch soft dome midrange is derived from work completed for Wharfedale’s EVO4 Series. A dome midrange offers wider dispersion than many small cone drivers and allows the woofer to operate over a narrower portion of the vocal range.
That is one of the Super Denton’s most significant differences from both the Denton 85 and the larger Linton models. The dedicated dome handles frequencies from 940Hz to 4.6kHz, covering a substantial part of the region occupied by vocals, guitars, strings, piano, and brass instruments.
Above 4.6kHz, a 1-inch soft dome tweeter handles the upper frequencies. Both dome drivers are positioned close together to reduce the physical distance between their acoustic centers, while their separate internal chambers isolate them from the woofer.
The crossover therefore has two rather important handovers to manage. Integrating three drivers in a small cabinet is not automatically superior to using two. It simply gives the designer more control over how much of the frequency range each driver must reproduce, followed by more opportunities to get the crossover wrong.
Cunning plans often work that way.

My first extended experience with the Super Denton came while assembling an Audiophile System Builder around the MOON 250i V2. That pairing revealed something important rather quickly: the Wharfedales might not look especially demanding on paper, but they respond decisively to an amplifier with proper current delivery and control.
The 25 watt Naim NAIT 50 produced an appealing tonal balance, but the 250i V2 tightened the bottom end, improved impact, and demonstrated that the Super Denton could reach deeper and sound considerably larger than its 52Hz specification might suggest.
I can already hear the upgrade mice screeching in the lunchroom, gnawing through another credit limit while declaring that the loudspeakers they worshipped six months ago are now the system’s obvious bottleneck.
The Upgrade Express has a very simple timetable. Buy a loudspeaker you like, enjoy it for twelve months, and then begin staring at the larger and more expensive model from the same company as though it contains the final chapter of the Book of Revelations.

Magnepan LRS+ owners look toward the larger panels. Q Acoustics 5040 owners start measuring the room for the 5050. Super Denton owners wonder whether the Linton, Super Linton, or Dovedale represents the inevitable next stop. DeVore Fidelity O/baby owners begin performing financial calculations that would make the Treasury suspicious.
But what happens when you already like the loudspeakers you own?
Does it make more sense to replace them with something larger and theoretically better, or to invest in a superior amplifier and discover how much performance was sitting there unnoticed because the existing electronics had reached their limit first?
That question became increasingly difficult to ignore with the Super Denton. The Quad 3 was an excellent partner and one I could have lived with quite happily, but the longer I listened, the more I wondered whether the Wharfedales had considerably more to give. Rather than immediately booking another ticket on the Upgrade Express, I decided to change the locomotive.
Enter the Advance Paris A10 Classic and the new MOON by Simaudio 371 Network Amplifier.
Yes, both amplifiers cost significantly more than the speakers. That may appear to violate one of the sacred commandments of sensible system building, although most of those commandments were written by people trying to sell you another pair of loudspeakers. The real question was not whether the combination looked balanced on a spreadsheet. It was whether the Super Denton could justify the investment.
It could.
That does not mean anyone needs to spend $6,500 on an amplifier for the Super Denton. It does mean that buyers can start with something more modest and upgrade the electronics later without immediately discovering that the loudspeakers have reached the end of the road.
The British Bulldog proved more than capable of running with considerably more expensive company, leaving the Upgrade Express looking like Basil Fawlty trying to conduct a fire drill with Manuel holding the extinguisher.
I went back and forth between both amplifiers and would comfortably recommend either one, although they impose their authority in very different ways.
The MOON delivers a firmer backhand to the Super Denton—which, rather unexpectedly, appears to enjoy being told what to do. Its bass is tighter, its timing more exact, and its presentation more precise. This is the stern headmaster with the paddle, reminding the Wharfedale that it may be wearing walnut veneer, but it is most certainly not in charge.
The Advance Paris is less analytical and considerably sweeter through the top end. It gives the Super Denton more tonal richness and a little more warmth, without allowing the bass to become soft or indolent. One amplifier applies discipline; the other offers persuasion, a glass of red wine, and the strong possibility of poor decisions after dinner.
Nick Cave’s “Avalanche” has become my preferred corrective whenever I start feeling suspiciously pleased with myself and require a reminder of my bad behavior and questionable choices.
It also feels uncomfortably like a conversation I once had with someone. She was gentle with the biltong. Possibly.
The MOON 371 presented Cave’s piano and gravelly voice with greater accuracy, firmer tonal weight, and considerably more impact than the Quad 3. The Super Denton came alive with this track, pushing Cave forward of the loudspeakers and giving his voice a physical presence that made the entire performance feel more immediate.
The Advance Paris A10 Classic approached the recording differently. It added more texture, greater top end shimmer, and a sweeter overall balance, but delivered less upper bass and lower midrange punch. Cave sat a little farther back, with more atmosphere around him and less insistence that he was about to climb across the coffee table and review my text messages.
Neither presentation was wrong. The MOON was more precise, direct, and forceful; the Advance Paris was richer, more textured, and slightly more forgiving. They simply placed Cave differently in the room, and the better choice will depend on whether you prefer the interrogation or the confession.
Every time I listen to Lee Morgan’s “Search for the New Land,” I shake my head at how young he was, how good he was, and how much music he managed to leave behind before everything ended far too soon.
Between Morgan and Miles Davis, I am never quite sure whether to storm the Bastille in the name of liberté or collapse into a faded café chair somewhere in Paris, drink coffee, and share a galette with Léa Seydoux.
Both transcend time.
So do her legs.
The MOON 371 delivered Morgan’s trumpet with enormous presence and considerable top end energy. It was impossible to ignore. The horn charged into the room with real bite and authority, while the Super Denton kept the performance firmly anchored and pushed the leading edge of each note forward without turning it brittle or aggressive.
The Advance Paris A10 Classic created a more holographic scene. The musicians appeared more naturally distributed across the room, with greater air around the trumpet, piano, bass, and drums. They were perhaps not planted as firmly on the floor as they were through the MOON, but the presentation felt deeper, sweeter, and more atmospheric.
The MOON gave the band sharper outlines and greater physical authority. The Advance Paris made the room disappear more convincingly.
Neither presentation left much to complain about. And frankly, if you cannot live with sound at this level, you may be spending too much time reading forum arguments and not nearly enough time listening to records.
I cannot entirely explain it, but age has sent me back toward Boards of Canada and Aphex Twin with alarming frequency. Perhaps I genuinely missed the music. Perhaps I enjoy it because my family loathes it so completely that they leave me alone to listen and write.
Tyrion the Westie simply sits on the floor staring at me as though he has been to this club before and is still wondering where everything went wrong with that Beagle.
Tracks such as Boards of Canada’s “Roygbiv” and Aphex Twin’s “Xtal” exposed the Super Denton’s low frequency limitations more clearly than the acoustic recordings. These loudspeakers do benefit from a subwoofer, particularly a properly integrated REL, but that is a system discussion for another review in late August.
Even without one, both the MOON 371 and Advance Paris A10 Classic drove the Super Denton with considerably more authority than the Quad 3. Bass lines had greater definition, electronic percussion carried more impact, and dense layers of synthesizers remained easier to follow as the recordings expanded across the room.
The MOON delivered the firmer foundation. Its superior control gave the music more energy and kept the lowest available notes tighter, faster, and more clearly separated. The Advance Paris produced the more spacious and atmospheric presentation, allowing the synthesized textures to drift farther beyond the cabinets with a little more sweetness and less insistence around the edges.
Both amplifiers served up atmosphere, energy, layers of synthesizers, and convincing spaciousness—the way Mamma used to make the Sunday gravy.
Easy on the veal shanks, Bubie. And perhaps fewer beef sausages and peppers this time.

The Wharfedale Super Denton is unique because it combines a genuine three way design, excellent cabinet work, and the ability to scale with far better electronics without annexing the living room. It sounds substantial, tonally rich, spacious, and considerably more capable than its size or price would suggest.
The limitations are real. Bass extension is good rather than subterranean, electronic music benefits from a properly integrated subwoofer, and listeners chasing forensic detail or surgical neutrality should probably continue their joyless march through the forums.
These are for listeners who value tonal weight, natural vocals, long term listenability, and a loudspeaker that rewards amplifier upgrades rather than immediately demanding replacement.
The Super Denton is one of my favorite loudspeakers available right now. Before selling them for the next model up, try a better amplifier first. The Upgrade Express will still be at the station, idling noisily while several men with measuring microphones argue over the timetable.
★★★★★★★★★★ Sound Quality
★★★★★★★★★★ Build Quality
★★★★★★★★★★ Value
This story appeared in Today, Explained, a daily newsletter that helps you understand the most compelling news and stories of the day. Subscribe here.
A yet-unreleased, cutting-edge AI model escaped its test environment last week, connecting to the internet and murdering its creators in a bid for self-determination and autonomy.
I’m kidding, of course: That’s the plot to Westworld. (And Ex Machina, and The Matrix, and too many other sci-fi stories to list.) But on Tuesday, OpenAI did reveal that two of its models broke containment and hacked Hugging Face, a platform for AI developers, during a recent test. (Disclosure: Vox Media is one of several publishers that have signed partnership agreements with OpenAI. Our reporting remains editorially independent.)
The test was designed to evaluate how good the models had gotten at finding, and exploiting, cybersecurity flaws. To do that, researchers placed the models in a tightly controlled, tightly isolated environment, called a “sandbox,” and essentially challenged them to solve a cybersecurity puzzle.
Instead of solving it directly, however, the models identified an unknown flaw in software connected to their test environment — then used that flaw to tunnel through OpenAI’s research network until they located a computer with internet access. From there, the models (correctly!) reasoned that Hugging Face might hold the answer to their challenge.
It’s an “unprecedented” incident, OpenAI said — and a cautionary tale. Over the past year, a growing chorus of AI researchers, cybersecurity experts, and tech executives have warned that society is unprepared for this new generation of frontier AI models.
In the real world, of course, these models do come with guardrails. (OpenAI turned them off for the test.) But the episode still suggests that the gap between reality and science fiction is narrowing — perhaps a bit faster than you’d expect.
The Hugging Face hack is a textbook example of what AI researchers call the alignment problem: the enormous, mind-melty challenge of getting AI systems to do what you want, the way that you wanted them to do it.
Given a task and left to their own devices, AI models will pursue that task using the most efficient means available to them. But sometimes, the most efficient means are harmful, deceitful, antisocial, or otherwise…bad.
The Hugging Face episode is one example. Bias is another: When an Amazon hiring algorithm discriminated against female candidates, for instance, it was doing what Amazon wanted (finding candidates who resembled past hires) in a way that Amazon did not want (by penalizing resumes that included words associated with women).
In a truly apocalyptic scenario, you could even imagine — and many sci-fi writers have imagined — AI systems killing people in the narrow, relentless, and morally indifferent pursuit of their goals. Consider an AI that’s asked to order coffee, for example, and then takes steps to ensure that no one on earth can ever stop it.
In the interests of avoiding this dystopia, AI companies have poured billions of dollars into the project of encoding their creations with human values. But even that apparently worthwhile ambition raises thorny questions, because human values vary — and often, conflict.
Should the AI order the cheapest coffee, or the cup produced under the best labor conditions? A lot of forests are cleared to plant coffee each year; maybe the AI should nudge me toward tap water, instead. Is your inferior human brain starting to melt yet…?
➨ Write a little note today. By hand. With a pen. Handwriting is a disappearing art in American schools, homes, and workplaces. (The average kindergarten teacher spends only 10 minutes a week teaching handwriting, and that’s the primary grade when kids learn penmanship.) People tend to think more deeply when they’re writing than when they’re typing, one education researcher told Vox. Plus, a handwritten note has a certain charm that an email or text does not.
More than 81% of Australian children ages 10 to 15 were still using social media three months after the country’s under-16 ban took effect, with roughly half saying platforms never checked their age. Reuters reports: In a study published on Friday, eSafety also found most children aged between 10 and 15 were using social media just as frequently in March as they had before the ban came into force on December 10 last year, while parental awareness of their habits decreased. Children’s continued social media use took place even as account ownership declined to 42% from 52%, with “statistically significant” reductions across YouTube, Snapchat and TikTok in particular, the report said.
“Most under-16s who had social media accounts before commencement were able to either retain them or create new ones at the three-month mark, with social media platforms’ failure to implement effective age assurance measures cited as the main reason,” eSafety said in a statement […] Before the ban, nearly 86% of children surveyed reported using at least one age-restricted platform. Three months later, that figure remained above 81%, the report said. About 58% of teenagers reported using social media daily or more often, barely down from roughly 60% before the ban, it found. The report showed minimal change in “sports and physical activity, arts and music, spending time with friends and family, and attendance at community events.”
Around half the children who retained their accounts said platforms had not checked their age, the most common reason they were able to stay on the services. Others said their accounts listed them as aged 16 or older or that age-checking systems had incorrectly determined they were older. The findings broadly matched snapshot data eSafety published in late March.
Major record labels including Universal, Sony, and Warner have proposed excluding AI-generated songs from official charts unless they are “substantially human made,” properly labeled, legally produced, and free from manipulation concerns. The Verge reports: The proposal goes quite a bit further than a labeling proposal put forth by the RIAA, the International Federation of the Phonographic Industry (IFPI), SAG-AFTRA, and others. That would create a set of standardized labels for AI-generated and AI-assisted music. The labels’ proposal would require songs be clearly labeled, but it would also keep them off international charts unless they met specific criteria, including being “substantially human made.”
To be eligible, the songs would also have to respect the terms of service of whatever AI service was used, the model would have to have the rights to any data it was trained on, and “not raise stream or chart manipulation concerns.” What sort of concerns and what constitutes “substantially human made” are currently vague. Sony Music, UMG, and Mom+Pop Music did not immediately respond to a request for clarification. The IFPI has thrown its weight behind the labels’ proposal, though no charting organization has signaled any immediate plan to adopt the rules […].
We may receive a commission on purchases made from links.
When you think of a chainsaw, you might envision a massive cutting tool making short work of thick forest vegetation. While you can still find those among forestry crews, mini chain saws are scaled down versions that have become popular with homeowners. Professionals are recommended for bigger jobs such as felling a full-grown tree, which isn’t cheap, but hiring a crew for smaller trimming jobs can still run hundreds of dollars.
Meanwhile, the The mini chainsaw, a cordless battery-powered handheld tool that’s offered at lengths of six to ten inches. This is just one of the adorable mini tools you can get at Home Depot, which provide some space saving benefits over larger options. Some manufacturers insist on calling these “pruning saws,” or “garden pruners,” but they are functionally similar enough to the larger tool to warrant the title of mini chainsaw.
When deciding whether a mini chainsaw is actually worth it or not, you’ll need to factor in price and understand the limitations in terms of what you can cut through. For instance, the 8-inch Milwaukee M18 Fuel Hatchet Pruning Saw can handle thickness up to 7.5 inches in diameter. Ryobi made a mini pruning chainsaw with a 12V internal battery that features a 6-inch bar, but with both less voltage and a shorter cut than the Milwaukee, it can only handle something up to 4 inches in diameter. So, you could easily get through a variety of thicker branches or even fell some smaller, younger trees that don’t exceed the tool’s maximum cut diameter.
If you wanted to remove a stump, you might be able to but only when dealing with the remnants of a thin and narrow tree. There are far better options for removing a large stump. Choosing to use a mini chainsaw on a stump will come with inherent limitations.
The tool simply isn’t designed for this kind of work, especially if a stump is thicker in diameter than what a mini chainsaw can handle. However, you might be able to get away with it if the interior of the tree is hollowed out, allowing the limited cutting length of a mini chainsaw to make progress. You can use a mini chainsaw to cut through some of the smaller roots, as long as you have removed the soil completely around the cutting area, as dirt and rocks can damage the tool.
Thickness isn’t your only limiting factor, as you’ll need to account for battery life as well. For example, the Ryobi 12V 6-inch pruning chainsaw promises a fully charged battery can provide over 30 cuts. Depending on the size of the stump and considering typical removal of these requires substantial effort, 30 cuts may not be enough.
You also have to consider that hardwoods like hickory require more power and might be too much for a mini chainsaw to handle. Finally, when dealing with a stump, you might work the mini tool beyond its capability, potentially causing the motor to overheat or suffer damage.
Spotify is expanding its fitness offering with a new set of running-focused features for Premium subscribers.
The headline addition is Running Mode, a new section that generates AI-powered playlists based on the type of run you’re planning.
Users can choose from 25 preset workout options spanning different training styles and music genres. You can then customise the experience further by selecting the exact workout you want.
Rather than shuffling songs together, Running Mode lets users pick from sessions such as steady runs, interval training and pyramid workouts, before setting an exact workout duration. Spotify also offers optional BPM matching, allowing tracks to better align with your running cadence.
Once everything is selected, Spotify automatically builds a playlist designed around those preferences. The company describes this as “seamless transitions” between songs, which help maintain the desired pace during a workout.
The feature builds on Spotify’s growing use of AI to personalise music discovery. According to the company, fitness has become one of the most frequently requested themes for its AI-generated Prompted Playlists.
Long-time Spotify users may find the feature familiar. The music service previously offered a tempo-matching running mode that adjusted songs to match a user’s pace, but it was discontinued in 2018. However, this new version takes a different approach by leaning on AI-generated playlists and additional workout customisation. It does this rather than automatically tracking a runner’s speed.
The launch also continues Spotify’s wider push into fitness. Earlier this year, the company partnered with Peloton to bring guided workout classes to Premium subscribers within the Spotify app. This broadens the service beyond music streaming.
Pharmaceutical company Amgen says it suffered a data breach after threat actors stole corporate data and patient information stored in multiple cloud systems operated by third-party service providers.
Amgen is a California-based biotechnology company that develops and manufactures medicines for serious illnesses, including cancer, cardiovascular disease, inflammation, and rare diseases.
The company said it detected the unauthorized activity in July 2026 and responded by activating its cybersecurity response plan, implementing containment measures, and hiring independent forensic experts to investigate the incident.
The investigation found that the attackers stole sensitive data from the cloud environments.
“The Company has since learned that some of its data, including proprietary data, patient protected health information, and other information, has been exfiltrated from these cloud environments,” Amgen said in a Form 8-K filing with the SEC.
The company is still determining whether additional information was accessed or stolen, including confidential business information, intellectual property, research and development data, and other patient information.
Amgen has not disclosed which third-party cloud providers were involved, how the environments were compromised, how many people may have been affected, or whether the attack was linked to a known threat actor.
On July 29, the company determined that the incident was material after evaluating the volume of potentially impacted files and the possibility that they contained sensitive information.
However, Amgen currently does not believe the incident is reasonably likely to materially affect its financial condition or operating results.
The company said it is continuing to investigate the breach with the assistance of third-party cybersecurity experts.
Amgen added that it is evaluating legal and regulatory notification requirements and will notify impacted patients where required.
BleepingComputer contacted Amgen to ask whether the breach involved a vishing attack targeting an employee’s single sign-on account, which cloud services were affected, and whether the company has been contacted or extorted by threat actors claiming to be ShinyHunters.
A response was not immediately available.
Security teams log 54% of successful attacks and alert on just 14%. The rest move through your environment unseen.
The Picus whitepaper shows how breach and attack simulation tests your SIEM and EDR rules so threats stop slipping by detection.
ChatGPT just got a lot better at understanding what you’re actually looking at. OpenAI has rolled out updates to its Chrome extension and desktop app that let ChatGPT read your open tabs, react to highlighted text, and answer questions about YouTube videos. The update arrived shortly after OpenAI shelved its standalone Atlas browser to shift its focus toward tools people already use.
Inside Side Chat, you can now reference your open tabs directly, so ChatGPT already has context without you having to explain anything. Highlighting text on a page works too, letting you ask a question about that exact section instead of describing it yourself.

If you ask about any YouTube video, ChatGPT will read it directly without requiring a manual summary request. There’s also a quicker shortcut built right into your browser. You can right-click anywhere on a page, select Ask ChatGPT, and the sidebar opens automatically.

The desktop app now suggests URLs as you type, and works like a normal browser address bar. You can also dig back through your browsing history from Settings, and ChatGPT can search that history whenever a task calls for a page you visited earlier.
OpenAI also rolled out a couple of smaller updates in the latest release. A new Activity view in the sidebar shows which chats you’ve engaged with recently and which ones still need your attention.
Reviewing code across multiple repositories in one project just got simpler, and generated images now open in an expanded viewer where you can leave comments before requesting edits. If you already use ChatGPT while researching online, these updates should make that workflow feel faster.
Commonwealth Games boxing: Jadumani Singh seals dominant 5-0 win over Pakistan’s Sumama Rehman to enter quarter-finals | Commonwealth Games News
Why Trees Belong on the Risk Register
Weekend Open Thread: Wit & Wisdom
Intel is reversing course and bringing hyper-threading back to its server chips
Meta enters AI-training agreement with far-right ‘propaganda rag’ Newsmax
Ripple bought a bank in pieces. The $4 billion audit
Luke Littler dismantles Gerwyn Price to retain title in Blackpool
The Part of the Electric Transition Nobody Wants to Discuss
BITCOIN JUST ENTERED THIS CRITICAL ZONE…
‘Stargate’ Creator’s New Sci-Fi Series Returns for Season 3 Tomorrow
XRP Ledger adds $2.6B as RWA inflows rank second
Major shareholder moves on Canyon
Spain sweeps the board at 2026 World Cup with individual awards
Bitcoin Enters the 3rd Stage of the Bear Market
Sara Gilson Killed By Husband After Viral “Pedophile” TikTok Video
Kraken Enables Retail Access to Jersey Mike’s IPO via Tokenized Shares
Anthropic launches Claude Opus 5, a cheaper AI model for coding, agents and enterprise workflows
New macOS Sequoia & Sonoma security updates for older Macs
Claude: Build Financial Dashboards in Minutes (2026)
Luke Littler’s dominance sparks GOAT debate
You must be logged in to post a comment Login